Class Activation Mapping in Explainable Computer Vision: A Method-Centered Review of CNN, Transformer, and Foundation-Model-Era Visual Explanations
2026-08-12 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial Intelligence
AI summaryⓘ
The authors review a group of methods called class activation mapping (CAM) that help explain how AI models identify important parts of images or data to make decisions. They collected and studied 57 papers since 2016, showing the field has grown from simple techniques to more complex ones that use multiple layers, tokens, and new types of models like transformers. The authors organize these methods by how they work and what they aim to explain, highlighting trends toward more detailed and comparative explanations. They also note that evaluating these methods is still inconsistent, and their review points out both what has been achieved and what challenges remain.
Class Activation MappingExplainable AIGradient-based MethodsTransformer ModelsWeakly Supervised LocalizationFoundation ModelsToken AttributionEvaluation MetricsHeatmapsConvolutional Neural Networks
Authors
AmirHossein Eshghi, Hamid Saadatfar, Seyyed Ali Hoseini, AmirMohsen Eshghi, Siavash Arjomand Bigdel
Abstract
Class activation mapping (CAM) is one of the most widely used visual explanation families in explainable artificial intelligence. Its purpose is intuitive: it converts internal model evidence into a heatmap that highlights the image regions, convolutional channels, tokens, or patches that support a target class or concept. Since the first CAM formulation in 2016, the field has moved far beyond global-average-pooled CNN classifiers. CAM-style methods now include gradient-based post-hoc explanations, gradient-free score and ablation methods, high-resolution upscaling, weakly supervised localization and segmentation, transformer token attribution, causal and debiasing methods, and foundation-model-era approaches that use CLIP, DINO, SAM, or feature-distribution comparisons. This review synthesizes a strict corpus of 57 method-centered papers published from 2016 onward. The paper develops a taxonomy that separates methods by attribution mechanism, architectural dependence, and evaluation objective. It then reviews gradient-based CAMs, recent and hybrid CAM-style methods, and model-based or architecture-aware methods. Across the corpus, the main trend is clear: the field is shifting from explaining one class score in one low-resolution CNN layer toward comparative, multi-layer, probabilistic, token-aware, and foundation-model-aware explanations. At the same time, evaluation remains fragmented. Faithfulness, localization, robustness, computational cost, and human trust are often measured with different protocols. The review therefore emphasizes not only what each method contributes, but also which gap it leaves open and which later methods attempt to close that gap.